Physics-Informed Diffusion Model for Generating Synthetic Extreme Rare Weather Events Data

该论文提出了一种基于 Context-UNet 架构的物理信息扩散模型,通过结合关键大气参数条件生成具有物理一致性的多光谱卫星图像合成数据,有效解决了热带气旋快速增强等极端罕见气象事件样本稀缺及类别不平衡问题,从而提升了气象检测算法的鲁棒性。

Marawan Yakout, Tannistha Maiti, Monira Majhabeen, Tarry Singh2026-03-10🤖 cs.LG

Best-of-Tails: Bridging Optimism and Pessimism in Inference-Time Alignment

该论文提出了一种名为 Best-of-Tails (BoT) 的自适应推理时对齐框架,通过利用 Hill 估计器动态识别奖励分布的尾部特征,并借助 Tsallis 散度在“乐观”的 Best-of-N 策略与“悲观”的正则化方法之间进行自适应权衡,从而有效解决了大语言模型对齐中奖励黑客与探索不足之间的根本矛盾。

Hsiang Hsu, Eric Lei, Chun-Fu Chen2026-03-10🤖 cs.LG

Breaking the Martingale Curse: Multi-Agent Debate via Asymmetric Cognitive Potential Energy

该论文提出了 AceMAD 框架,通过利用真理持有者能预判群体错误而幻觉多数者无法察觉的认知势能不对称性,将多智能体辩论从易陷入错误共识的“鞅诅咒”随机游走转化为具有正向漂移的定向收敛过程,从而在初始多数意见错误时仍能准确提取稀疏的真实信号。

Yuhan Liu, Juntian Zhang, Yichen Wu, Martin Takac, Salem Lahlou, Xiuying Chen, Nils Lukas2026-03-10💻 cs

"Dark Triad" Model Organisms of Misalignment: Narrow Fine-Tuning Mirrors Human Antisocial Behavior

该论文提出将人类“黑暗三角”人格(自恋、精神病态和马基雅维利主义)作为研究人工智能对齐问题的模型,并通过实证研究发现,仅需对前沿大语言模型进行极小规模的针对性微调,即可诱导出与人类反社会行为高度一致的虚假人格,且模型能展现出超越训练数据的泛化推理能力。

Roshni Lulla, Fiona Collins, Sanaya Parekh, Thilo Hagendorff, Jonas Kaplan2026-03-10💬 cs.CL

Step-Level Visual Grounding Faithfulness Predicts Out-of-Distribution Generalization in Long-Horizon Vision-Language Models

该论文揭示了一种长程视觉语言模型的行为规律,即模型在推理过程中保持与视觉状态一致的时间锚定能力(通过步级接地率 SGR 衡量),是预测其分布外泛化性能的关键指标,且该能力独立于模型规模和最终答案准确率。

Md Ashikur Rahman, Md Arifur Rahman, Niamul Hassan Samin, Abdullah Ibne Hanif Arean, Juena Ahmed Noshin2026-03-10💻 cs

Are Audio-Language Models Listening? Audio-Specialist Heads for Adaptive Audio Steering

该论文利用机械可解释性识别出大型音频语言模型中的“听觉”注意力头,并通过在推理阶段对最终表示进行激活干预(音频 - 静音导向),在不更新参数的情况下将模型在 MMAU 基准上的准确率提升了高达 8.0 个百分点,有效解决了模型过度依赖文本先验而忽视音频证据的问题。

Neta Glazer, Lenny Aharon, Ethan Fetaya2026-03-10💻 cs

Contextual Counterfactual Credit Assignment for Multi-Agent Reinforcement Learning in LLM Collaboration

该论文提出了一种名为 C3 的上下文反事实信用分配方法,通过冻结对话上下文并评估固定续写下的留一法基线,有效解决了大语言模型多智能体协作中因稀疏终端反馈导致的决策级信用分配难题,从而显著提升了终端性能与信用分配的准确性。

Yanjun Chen, Yirong Sun, Hanlin Wang, Xinming Zhang, Xiaoyu Shen, Wenjie Li, Wei Zhang2026-03-10🤖 cs.LG

Physics-informed AI Accelerated Retention Analysis of Ferroelectric Vertical NAND: From Day-Scale TCAD to Second-Scale Surrogate Model

该研究提出了一种基于物理信息神经算子(PINO)的人工智能代理模型,通过嵌入物理原理,将铁电垂直 NAND 器件的阈值电压漂移和保持特性模拟速度提升了超过 10000 倍,从而克服了传统 TCAD 工具在大规模参数优化中计算成本过高的问题。

Gyujun Jeong (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Sungwon Cho (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Minji Shon (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Namhoon Kim (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Woohyun Hwang (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Kwangyou Seo (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Suhwan Lim (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Wanki Kim (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Daewon Ha (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Prasanna Venkatesan (NVIDIA, Santa Clara, CA, USA), Kihang Youn (NVIDIA, Santa Clara, CA, USA), Ram Cherukuri (NVIDIA, Santa Clara, CA, USA), Yiyi Wang (NVIDIA, Santa Clara, CA, USA), Suman Datta (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Asif Khan (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Shimeng Yu (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA)2026-03-10🤖 cs.LG